Sensor signals for machine tool and process health assessment.

During the current experimental testing, sensor data was collected to assess the condition of a machine tool via a 'fingerprint routine' that could be run at regular intervals, and a milling machining process of an aluminium workpiece. Physically simulated faults and errors were introduced to detect these in the collected signals. Machine tool failure modes: Heavy tool– Load a significantly heavier tool than the baseline tool. Unbalanced– Load a tool that has a lower balancing classification than the baseline tool. Feedrate-adjusted– Conduct the fingerprint routine with a set of marginally reduced feed rate and spindle speed overrides (corresponding to an even spread of 6-10% reduction). Machining process failure modes: Misalignment – Tilt machine tool’s bed by A: 0.27°, B: 0.27°, C: 0.32°. Surface cracks – Drill 1.84mm diameter bores into the part, on the cutting path, before recorded trials. Tool wear – Wear the cutting tool severely before recorded trials. The machining trials consisted of straight up-milling face cuts on 24 workpieces of aluminium with dimensions 200 x 120 x 85 mm held on a LANG vice inside a DMG Mori DMU 40 eVo linear CNC 5-axis milling machine. The ‘fingerprint routine’ consisted of isolated and combined movements of the X-, Y-, and Z-axes, as well as rotation of the spindle. Further details about the experimental procedures and the research can be found in the following publication: "The application of machine learning to sensor signals for machine tool and process health assessment", https://doi.org/10.1177/0954405420960892